Does nsd still zombie on 2.4 kernels when a handle goes stale? The traditional way to avoid the zombie problem is to set the lifetime of database handles to a very large value so once connected to Oracle they'll never be allowed to die.
I think this was solved a while ago. I just tried it earlier today (setting MaxIdle and MaxOpen to nothing and left my server running for a couple of hours. I then set MaxIdle and MaxOpen to 100 (100 seconds, I presume) with no problems.
I'm running Red Hat Linux 7.1, kernel 2.4.12, AOLserver 3.3.1+ad13, ArsDigita Oracle8 database driver v2.6, Oracle 184.108.40.206.1. I'm not running the ACS. Naturally, I ran a couple of database-backed Tcl pages (table-test.tcl, et al) before letting the server sit around. I haven't seen the stale db handle snafu nor any mention of it for many, many months.
I believe the problem lies elsewhere (i.e., not related to stale db handles) assuming that Simon is using current software (kernel, nsd, Oracle driver, Oracle).